نبذة مختصرة : Unreserved criticism and learning from mistakes is, according to Karl Popper, part of the scientific discourse required to ensure scientific endeavour. The trend towards automated data analysis seems to take this requirement into account. However, what is the case if the results of the data analytics rely on automated data collection which is riddled with trivial errors? Terje Tüür-Fröhlich pursues this question in her dissertation, which has now been published in the series "Schriften zur Informationswissenschaft" published by the Hochschulverband für Informationswissenschaft (HI). On the basis of thorough and systematic literature research, Tüür-Fröhlich documents each step of her research. In a separate chapter, she gives valuable insight into the genesis of Citation Indexing on the basis of a content analysis of the correspondence between Joshua Lederberg and Eugene Garfield. Using case study as method, the author highlights the effects of "not deleted, not shared or retracted errors", which not only determine the careers of scientists, but also the scientific communication and evaluation of results. For this detailed study, which is the bulk of the work discussed here, the author focuses on the Social Sciences Citation Index (SSCI) of the database provider Thomson Reuters. Tüür-Fröhlich admitted the research was limited due to the data access restriction which is more than ever dependent on an institutional affiliation. The results are presented as qualitative and quantitative case studies.The core of the investigation, which is based on innovative methods like "ping-pong method" and „chain referral sampling“ for "Pierre Bourdieu", lead to n = 85 mutated name variants and missings for Pierre Bourdieu in the SSCI. The case studies read like a mix of detective novel and science fiction. The works of one author contain error phenomena for which Tüür-Fröhlich develops her own terminology. Examples are "UBO" for "unidentifiable bibliographic object" or "Zombie References", which are explained in a chapter heading as "Anything Is Wrong But Something Has Been Indexed". The conclusions of the study clearly state that all these "trivial" mistakes distort the citation counting and thus affect the impact factor - this not only applies to the authors, but also to their institutions, research fields and countries. Summarising, this book clearly shows how urgently qualitative methods are needed - at least in the evaluation of scientific performance as well as in science itself - as a supplement to quantitative methods. Some of the errors analysed are beyond the scope of quantitative research, which in addition does not take into account the genesis of the errors - and thus does not contribute to minimizing errors. The present book is highly recommended for data scientists, librarians and provides good impulses for a critical examination for database research and scientific knowledge processes.
Processing Request
No Comments.